Accuracy Of Smart Watches In Detecting Heart Rate During Different Exercise Conditions

Smart watches (SW) have gained popularity for monitoring heart rate (HR) during daily activities and exercise. Rapid advancements in technology, including both hardware and software updates, necessitates an understanding of the accuracy of current devices and their derived metrics. Furthermore, SWs determine HR indirectly, which introduces the possibility of influence by external factors such as exercise type and intensity. PURPOSE: Evaluate the accuracy of commercially available SWs in detecting HR compared to ECG measurements across various conditions. METHODS: 169 participants (83F/86M; age: 31.9 ± 10.9 yrs; ht: 172 ± 9.95 cm; wt: 70.6 ± 12.5 kg) completed a single lab visit with three exercise trials: Treadmill (TR, 3 min stages: Rest, 4.8kph, 9kph, 12kph, 15kph); Treadmill Interval (INT, 3 repeating stages: 2 min 4.8kph and 1 min 12kph); and Cycle Ramp (CYC, 3 min stages: 50 W, 100 W, 150 W, 200 W). During each trial, participants wore 2 different consumer SWs (SW1 and SW2) on randomized wrists and were outfitted with 12-lead ECG for HR monitoring. HR from each SW was time synced with ECG and mean absolute percent error (MAPE) was calculated across trials for each SW compared to ECG data. Paired t-tests were performed on MAPE values between SW1 and SW2 for each trial. RM ANOVAs were used to assess differences within each device between the three different trials. RESULTS: SW2 reported significantly less error than SW1 in TR, INT, and CYC (p < 0.001). Both SW1 and SW2 reported significantly lower error during CYC compared to INT (SW1: p = 0.02; SW2: p < 0.001) and TR (SW1: p = 0.05; SW2: p = 0.01; Figure) Conclusions: Overall, consumer SWs demonstrate low error in HR measurements (MAPE range 1.1-3.6%) compared to ECG during a range of exercise type and intensity. SW1 provides significantly lower error in HR metrics compared to SW2. Both devices reported less error during CYC, likely resulting from differences in arm motion. Supported by: This study received funding from Samsung Electronics Co., Ltd.